fix(podman): resolve macOS Podman socket dynamically - #3135
Conversation
On macOS, Homebrew-installed Podman does not create the default socket path that the Podman driver probes. Document the OPENSHELL_PODMAN_SOCKET override and the podman machine inspect lookup in both the compute drivers reference and the debug-openshell-cluster skill. Fixes NVIDIA#1690 Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
|
All contributors have signed the DCO ✍️ ✅ |
|
I have read the DCO document and I hereby sign the DCO. |
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
johntmyers
left a comment
johntmyers
left a comment
There was a problem hiding this comment.
gator-agent
PR Review Status
The macOS Podman fix is project-valid and the focused socket changes are ready to continue, but the latest commit also removes a repository-advertised troubleshooting skill while leaving its callers and documentation intact.
Action required: restore .agents/skills/debug-openshell-cluster/SKILL.md, or move an intentional coordinated removal into a complete, separately scoped change.
Blocking findings:
GATOR-b92131c8-01: the deleted skill remains referenced by repository guidance and companion workflows.
Carried findings:
- None
Gator metadata
- Validation: Concentrated macOS Podman compatibility fix linked to triaged issue #1690
- Docs: Fern compute-driver reference updated; existing reference-folder navigation covers the page
- Checks: Current-head branch and Helm gates are waiting for the copy-PR mirror
- E2E: Required for Podman runtime behavior; dispatch deferred until blocking review feedback is resolved
- Head SHA:
b92131c8a572bf935fd60a0f2784116b823ec676 - Base SHA:
8bc79552638a625971f8c6d30c26efa917a19cf9 - Merge base SHA:
06f0aa6fe30177f2aecba9af396649adae4d7cdd - Patch ID:
4bae9c48f63693d1ea65d2a789486c786b681b6d - Gator payload:
8 - Review mode:
initial - Previous reviewed SHA: none
- Review budget exhausted: no
- Maintainer decision required: no
- Next state:
gator:in-review
johntmyers
added
the
gator:in-review
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
|
Label |
|
/ok to test 69bf396 |
johntmyers
left a comment
johntmyers
left a comment
There was a problem hiding this comment.
gator-agent
PR Review Status
The author-only follow-up resolves GATOR-b92131c8-01: the effective PR tree keeps the canonical public troubleshooting skill and applies the intended macOS Podman guidance there. The bounded delta review found no new blocking findings, and the required Branch Checks, Helm Lint, and E2E workflows are queued for the current head.
Blocking findings:
- No blocking findings remain
Carried findings:
- None;
GATOR-b92131c8-01is resolved
Gator metadata
- Validation: Concentrated macOS Podman compatibility fix linked to triaged issue #1690
- Docs: Fern compute-driver reference and canonical public troubleshooting skill are updated; existing reference navigation covers the page
- Checks: Current-head Branch Checks and Helm Lint workflows are queued
- E2E:
test:e2eapplied;/ok to test 69bf396d84f2b52f2b96b5c4e24e237211913560posted; current-head Branch E2E workflow is queued - Head SHA:
69bf396d84f2b52f2b96b5c4e24e237211913560 - Base SHA:
8bc79552638a625971f8c6d30c26efa917a19cf9 - Merge base SHA:
06f0aa6fe30177f2aecba9af396649adae4d7cdd - Patch ID:
a80112376cbab89bacc549dc61b77904c9dfd8ba - Gator payload:
8 - Review mode:
follow_up - Previous reviewed SHA:
b92131c8a572bf935fd60a0f2784116b823ec676 - Review budget exhausted: no
- Maintainer decision required: no
- Next state:
gator:watch-pipeline
johntmyers
added
gator:watch-pipeline
elezar
left a comment
elezar
left a comment
There was a problem hiding this comment.
Can we remove the unrelated skills changes from this PR and keep it focusses on the macos-specific podman machine changes.
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
johntmyers
left a comment
johntmyers
left a comment
There was a problem hiding this comment.
gator-agent
PR Review Status
Thanks @elezar. I checked your request to keep this focused on the macOS Podman Machine change against the two author commits and the effective patch after the merge from main. The unrelated E2E shell edits net out against the previously reviewed head, and the bounded delta review found no new blocking code defects, but skills/debug-openshell-cluster/SKILL.md still has the nine-line addition you asked to remove.
Action required: @gmenher, remove the remaining skills/debug-openshell-cluster/SKILL.md change, or obtain an explicit maintainer decision to retain it in this PR.
Blocking findings:
- No new Gator blocking findings
Carried feedback:
- @elezar's requested scope change remains unresolved
Gator metadata
- Validation: Concentrated macOS Podman compatibility fix linked to validated issue #1690
- Docs: Fern compute-driver reference is updated; maintainer-requested skill scope cleanup remains
- Checks: Current-head required checks are pending; pipeline handoff is deferred until review feedback is resolved
- E2E:
test:e2eremains required; current-head dispatch is deferred until review feedback is resolved - Head SHA:
39f5ede5df3633a5e785254457dc61d38591855a - Base SHA:
0f0c94bda3a2056bd3413accd13e361488288cc4 - Merge base SHA:
0f0c94bda3a2056bd3413accd13e361488288cc4 - Patch ID:
a7eba7991fb1b953ad1fd0497a1bceef33f91f79 - Gator payload:
8 - Review mode:
follow_up - Previous reviewed SHA:
69bf396d84f2b52f2b96b5c4e24e237211913560 - Review budget exhausted: no
- Maintainer decision required: no
- Next state:
gator:in-review
johntmyers
added
gator:in-review
3 participants
Summary
This PR removes the last macOS Podman socket assumption from the driver-config volume e2e path and makes the Podman-backed e2e harness respect the socket reported by
podman machine inspect.On newer macOS Podman setups, the legacy
~/.local/share/containers/podman/machine/podman.sockpath is not a reliable source of truth after restartingpodman machine. That meant the driver-config volume helper and the Podman e2e flow could fail even when the VM was healthy and reachable through the socket Podman itself reports.Related Issue
Closes #1690
Changes
OPENSHELL_PODMAN_SOCKETwhen it is already set, so explicit user overrides still win.podman machine inspectbefore falling back to the legacy socket path.CONTAINER_HOSTwhenever the Podman socket is resolved so child processes and follow-on commands use the same endpoint.podman machine inspectoutput.While validating the macOS Podman Machine change, I hit two unrelated harness bugs in gateway-common.sh and stage-prebuilt-binaries.sh caused by set -u plus empty arrays. I included those fixes because they blocked local e2e verification, but they are separate from the macOS socket change.
Testing
mise run pre-commitpassesChecklist